Study On Multidimensional Classification Method Of User Behavior In Smart Grid And Its Application

With the comprehensive construction of smart grid, deep integration of information and communication technology, the level of intelligence and interaction in intelligent electric field significantly improve, and adding massive increase in power grid elements, such as distributed power, energy storage devices, electric vehicles, micro grid, make the user’s behavior to diversified development. And with the power grid reform, the release of the electricity side makes the user’s position significantly improved, which requires the sale of electricity companies to fully understand the behavior of users to enhance their ability to compete for the user’s resources. Therefore, it is necessary to study user behavior in all aspects through the multiple dimensions of customer segmentation. It is very important to meet the different types of users for personalized value-added services and better realize the interaction between the’person and power grid’.Firstly, this paper analyzes the defects of the existing user classification method, proposes the user classification rules. Based on the user’s characteristics, the’evolution’ principal component analysis method to classify the users with different types of load patterns is proposed. Then, the typical daily load curve of each user is extracted by the cosine similarity theorem. The results show that the proposed method has a significant effect on the classification of users with similar load patterns. Finally, according to the classification rules, the classification rules based on the user’s power characteristics are realized.Secondly, using SAS data analysis software to study the user’s credit risk assessment method, through the SAS Enterprise Miner for modeling, the decision tree model, regression model and neural network model used for user credit risk assessment are compared through three different training data sets and using ROC, LIFT chart to determine the accuracy of the three models. The experimental results prove that the accuracy of the model will be different according to the number of enterprises’ marketing. Therefore, we need to evaluate the credit risk of the user according to the model with the highest accuracy in different scenarios.At last, because of the uncertainty of the user’s daily use, the user’s daily consumption of electricity and electricity load curve are different, so it is one-sided to determine the potential of the use of electric power strategy simply based. on consumption characteristics, credit and power consumption. This paper uses fuzzy comprehensive evaluation method to study the potential of the user’s power strategy guidance from the three dimensions of user’s power consumption with consideration on the uncertainty of each user’s daily use of electricity. Finally, we get the potential value of each user’s power strategy to achieve a real sense of customer segmentation, while providing intuitive data support for the sale of electricity companies guiding the user’s power consumption.